Dear colleagues

I have a question about using Statistic Parametric Mapping (SPM) methods on biomechanics time series data.

I want to use this method to compare kinematics time series data between two different groups. When I run SPM on 100 data points, there are no significant differences in any phases of the gait cycle. However, when I split the time series to ten parts with 10 data points in each part and run SPM on each split phase, I can see significant differences in some phases.

Which model is correct?